Lecture 3

Question 1

If a regression model has nonnormally distributed errors, what is the effect of the unbiasedness of OLS?

Answer 1

Question 2

If a regression model has nonnormally distributed errors, what is the effect of the consistancy of OLS?

Answer 2

Question 3

If a regression model has nonnormally distributed errors, what is the effect of BLUEness of OLS?

Answer 3

Question 4

If a regression model has nonnormally distributed errors, what is the effect of the asymptotic efficiency of OLS?

Answer 4

Question 5

If a regression model has nonnormally distributed errors, what is the effect of the variance of OLS?

Answer 5

Question 6

If a regression model has nonnormally distributed errors, what is the effect on the distributions on the t-test or F-test of OLS coefficients? (for finite observations)

Answer 6

Question 7

If a regression model has nonnormally distributed errors, what is the effect on the distributions on the t-test or F-test of OLS coefficients? (for infinite observations, i.e., large n)

Answer 7

Question 8

What three properties of OLS are different if the errors are nonnormally distributed?

Answer 8

Homoskedatic errors are non-normally distributed, OLS is not efficient, t-statistic and F-statistic do not have their respective distributions.

Question 9

Answer 9

Question 10

What are three tests that can be used to test normality?

Answer 10

Question 11

Answer 11

WERK DEZE UIT

Question 12

Answer 12

Question 13

Answer 13

A Uniform [0, 1] distribution

Question 14

Answer 14

Question 15

Answer 15

We test H₀: γ_i = 0 (etc.) We test with ɑ = 0.05/n

Question 16

Answer 16

Question 17

Answer 17

Test if 𝛿 = 0 with wald test

Question 18

What is a disadvantage of the RESET test?

Answer 18

Question 19

Answer 19

Question 20

Answer 20

Here we can test whether intercept and slopes are same for both groups:

H₀ :δ₀ = δ₁ =δ₂ = 0. The F-test for this is the Chow (break)test.

Question 21

Answer 21